Particle.news

Download on the App Store

Google Inks First Machine Learning Demand-Response Deals With US Utilities

These agreements let Google shift non-urgent ML tasks during periods of high demand to bolster grid reliability as part of its 24/7 carbon-free energy strategy.

Image art by Paul Gerke via ChatGPT-4o.
Image
Image: Romolo Tavani, Getty Images/iStockphoto
Image

Overview

  • Google has signed its first formal demand-response agreements with Indiana Michigan Power and the Tennessee Valley Authority to curtail AI data center power consumption during grid peak events.
  • The pioneering pacts specifically target machine learning workloads, allowing Google to shift or reduce non-urgent compute tasks when utilities request relief.
  • This move expands on a 2023 pilot with Omaha Public Power District where Google curtailed ML workloads during three separate grid strain events.
  • Utility leaders say Google’s load flexibility can help manage generation and transmission resources more effectively, potentially reducing the need for new infrastructure.
  • Integrating demand-response into Google’s 24/7 carbon-free energy ambition underscores its effort to balance rapid AI growth with grid stability and consumer cost management.